Browse Books

Reviewer: Richard L. Frautschi

In a richly documented study, Steedman seeks to develop a “principled theory of natural grammar” (p. xi) compatible both with syntactic order/disorder constituency, and with psychological and computational mechanisms that can link surface forms to interpretable meaning representations. Without breaking rank with Chomsky, who remains an unimpeachable master, Steedman nonetheless gently nudges the envelope by proposing new, and perhaps unorthodox, strategies as complements to conventional notions of grammatical universals. With occasional wry asides, such as Sapir's remark that “all grammars leak,” Parts 1 and 2 deal with treatments of theoretical and empirical “competence” subsumed under the unifying aegis of the author's “strict competence hypothesis.” Steedman argues that surface syntax analysis should respect the logic of monotonic-monostratic formalisms. After numerous examples of transformation practices, he cautiously proposes an enhanced “combinatorial categorical grammar,” folding additional components (including intonation, theme and rheme, and discourse analysis) into case studies as aids to deep structure mapping. Part 3 dwells briefly on parser architecture, whether human or computational. Steedman notes that artificial languages are presumed to have strong rule-to-rule bonds between semantics and syntax, whereas natural languages display a “dysfunctional” relationship between mapping and interpretation. New relationships between grammars and processors are suggested fleetingly. In reading his proposed enlargements to the fortress of generative theory, I could not ignore the author's unshakable trust in declaratizable components: utterance, by all means, over silence. Yet, recurring allusions to the unsystematic acquisition of semantic and syntactic “concepts” in children, as well as our near ignorance of neural and social constraints on the evolution of natural languages, remain unsolved mysteries, a messiness to be kept outside of the protective walls of Urgrammatik. While rightfully concerned with the dangers of proliferation in adaptations of an enhanced standard theory, especially when applied to programming languages, Steedman asks whether another theory with fewer assumptions might be possible, especially since none of the present theories of grammar are descriptively adequate. Finally, a reconciliation of tensions between opposing epistemologies—movement (for example, dynamism) versus immutability (for example, truth values), which the present enhanced theory presumably seeks to accommodate—fails to materialize in the last chapter, if only because of the author's stronger faith in a linguistic tradition of short string analysis. Despite a few glimmers of polytonicity and the obligatory call for more research, a final caveat warns against an “evolutionary Panglossism” (p. 261).